Review by Exception for Electronic Batch Records: What Has to Be True First

Ask five plants what “review by exception” means and you get five answers, from “QA skims the exception list” to “nobody reads the record unless the system shouts.” Only one of those survives an inspection. The idea is sound: a reviewer who must read 150 pages of a paper batch record to find the one transcription error is working against human attention, not with it. But review by exception (RBE) is a claim you have to earn with evidence, and the evidence is mostly about the process, not the software.

This is what an electronic batch record review by exception needs before a QA head or QP can defend it, what the primary texts do and do not say, and where we see plants get caught out.

Paper BMR versus eBMR: what the reviewer is really doing

On a paper batch manufacturing record, the reviewer is the control. Every check that matters (did the operator weigh within tolerance, was the hold time met, was the second signature applied, does the yield reconcile) is found, or missed, by a person turning pages after the batch has finished. The record is passive. A missing initial is discovered days later, when the person who should have written it is on another shift and memory has faded.

A properly built electronic batch record (eBMR) moves the checks to the moment of entry. The system compares the captured value with its limits, blocks a step that is out of sequence, refuses a signature from someone whose training has lapsed, and flags an absent entry. The record becomes active. That shift is what makes RBE possible at all: if the system has already tested every value against its limit at write time, the reviewer’s time is better spent on the cases where the test failed. If the eBMR is only a typed copy of the paper form, with free-text fields and no enforced limits, there is nothing to review by exception, and you should not pretend otherwise.

What the regulations actually say

Here is the point that surprises people: the phrase “review by exception” does not appear in current EU GMP. The legal footing is narrower and more conditional than vendors tend to suggest.

EU GMP Chapter 4

Paragraph 4.20 lists what a Batch Processing Record must contain, from batch number and dates through in-process controls, yields, and notes on special problems with signed authorisation for any deviation. Directly after that list, the chapter adds a note: “Where a validated process is continuously monitored and controlled, then automatically generated reports may be limited to compliance summaries and exception/out-of-specification (OOS) data reports.” (Chapter 4). That note is the basis for RBE. Read it again: it has two conditions (validated, and continuously monitored and controlled) and it speaks of the automatically generated reports, not of QA’s responsibility for the record as a whole.

EU GMP Annex 15

Annex 15 supplies the second leg, because “validated” and “continuously monitored” are not claims you can make by assertion. Section 5 covers process validation, and its ongoing process verification paragraphs say manufacturers “should monitor product quality to ensure that a state of control is maintained throughout the product lifecycle with the relevant process trends evaluated” (5.29), under an approved protocol with a report and statistical tools where appropriate (5.31), and that this verification supports the validated status documented in the Product Quality Review (5.32) (Annex 15).

ICH Q10

ICH Q10 section 3.2.1 asks for a process performance and product quality monitoring system that uses the control strategy, analyses parameters “to verify continued operation within a state of control,” and feeds in complaints, rejections, deviations and audit findings (ICH Q10). It describes the monitoring you must have. It does not tell you how much of the batch record you may stop reading.

Annex 11

Annex 11 section 9 expects, based on risk, a system-generated audit trail of GMP-relevant changes and deletions, with the reason documented, and says audit trails should be “regularly reviewed” (Annex 11). Switching to RBE does not remove that duty. It makes it more visible, because the audit trail is where an unwanted change hides when nobody reads the full record.

Draft revisions. The European Commission ran a stakeholder consultation on a revised Chapter 4, a revised Annex 11 and a new Annex 22 on AI. The documents are drafts. The consultation opened on 7 July 2025 and closed on 7 October 2025 (consultation page). As of 11 November 2026, the Commission’s Volume 4 page still lists the January 2011 versions of Chapter 4 and Annex 11 as the current texts (EudraLex Volume 4), and we found no adopted revision. Treat anything you read about the drafts as consultation material, and do not write them into an SOP as requirements.

The US view: 21 CFR 211.188 and 211.192

The FDA text is blunter. Section 211.188 says batch production and control records must include documentation that each significant step was accomplished, including dates, equipment, component identities, weights, in-process and laboratory results, yields, and the identity of the people performing and checking each step. Where a significant step is done by automated equipment, the record must identify the person checking it. It also includes any investigation made under 211.192.

Section 211.192 then says that “all drug product production and control records, including those for packaging and labeling, shall be reviewed and approved by the quality control unit to determine compliance with all established, approved written procedures before a batch is released or distributed.” It requires any unexplained discrepancy, including a yield outside the established percentages, or any failure to meet a specification, to be thoroughly investigated, and the investigation to extend to other batches that may be involved.

There is no exception wording in either section. In our reading, an RBE procedure can still be how the quality unit performs that review, provided it can show the system has verified the records it does not read line by line, and that the records are reviewed and approved. That is an argument you have to make and document. It is not a safe harbour. In our experience, many sites that supply the US keep full review as the default and treat RBE as a per-product, per-market permission.

Annex 16: the QP certifies, whichever way the record was reviewed

RBE changes how records are reviewed. It does not change who releases. Under Annex 16, batches “should only be released for sale or supply to the market after certification by a QP,” and until then the batch stays at the manufacturing site or ships under quarantine to an approved site (4.1). Among the QP’s responsibilities: that all records are complete and endorsed by appropriate personnel, with all required in-process controls and checks made (1.7.11), that all manufacturing and testing processes remain in the validated state (1.7.12), and that all investigations pertaining to the batch, including out-of-specification and out-of-trend ones, have been completed to a sufficient level to support certification (1.7.16) (Annex 16).

Certification is the release. So the sequence that makes sense is QA review and disposition first, with the review path recorded, and then the QP certification as a separate act by a named person. A QP who is asked to certify a batch released “by exception” must be able to see which path was used and why it was allowed for that product.

What has to be true before RBE is defensible

1. The process is validated, and the evidence is current. Not “we ran three batches in 2019.” Ongoing process verification is running, the Product Quality Review has recent conclusions, and changes since validation went through change control and were assessed for revalidation.
2. Critical process parameters are identified and enforced, not just recorded. The system must know the limits and act on them. If a CPP can be typed in free text, it cannot be reviewed by exception.
3. The exception rules are written down, versioned and validated. Which events count as an exception, how each is graded, where it goes next, and which version of the rule set classified a given batch. “The vendor configured it” is not an answer.
4. Absence counts as an exception. A missing mandatory entry, a step with no execution record, a signature never applied, a laboratory result never received. If the rule set only detects bad values and says nothing about missing ones, the exception list looks clean precisely when the record is incomplete.
5. Audit-trail review is defined and performed. By someone independent of the activity, recorded, scoped to the batch, and covering the changes to critical data.
6. The decision is per product, with an owner. A named person signs that this product, in this market, is eligible, and the decision can be withdrawn.

How exceptions should be graded

Grading is a risk decision and it should look like one. We use four tiers because they map onto four different reviewers:

• Critical: a confirmed out-of-specification result, a bypassed control, a reconciliation breach, an operator with expired training. These open a deviation and block release until it is dispositioned.
• Major: an out-of-sequence step, a missing mandatory step or signature, a timing window exceeded, an active equipment alarm. These go to QA review.
• Minor: a late entry, a correction, a value inside specification but past the alert limit. A supervisor can review and close with a stated reason.
• Info: visibility only, for example an entry captured offline and synced later. It is listed, and nobody has to act.

The reasons matter more than the labels. Write the rationale for each rule’s grade into the validation file, so that when an inspector asks why a late entry is minor and a missing signature is major, the answer is on paper before the question is asked.

What QA still reviews in full

RBE narrows the routine reading. It does not remove these:

• Every critical and major exception, with the investigation behind it and the conclusion.
• Open and closed deviations linked to the batch, and the assessment of whether they touch other batches (211.192 asks for this explicitly).
• Laboratory results against the release specification, and any OOS investigation.
• The audit-trail review for the batch and its outcome.
• Completeness: every mandatory step and signature accounted for.
• Changes to the master record or process since the last batch, if any.
• A positive statement by the reviewer of whether issues were found. A blank is not the same as “none.”

Inspection pitfalls we see

Rules too loose. The exception list is short because the thresholds were set wide, not because the process is stable. An inspector who compares your exception rate with your deviation and complaint history will notice a plant with few exceptions and many complaints.

Unvalidated limits. Alert and action limits that were copied from the paper BMR, or tightened and loosened by an administrator without change control. If the limit that triggers an exception is not itself justified, the whole concept rests on sand.

Exceptions closed without justification. “Reviewed, OK” against a major exception tells the inspector the reviewer clicked through. Every closure needs a reason that someone else could challenge.

RBE applied by default. The permission should exist per product and market, backed by validation status, and the fallback should be full review. Plants that turn RBE on globally, because the software offers it, get asked which validation report supports each product.

How Corpex-EBR handles it

In Corpex-EBR, review by exception is a permission, not a mode. A policy record per product and market must say it is permitted, that the process is validated, and that it is continuously monitored; with no record, or any one of those false, the release is refused and full review is required. A change in policy is a new row, so the history of what was permitted when is kept. The disposition stores whether it went by the full or the exception path. Exceptions are raised at write time by 24 rules, each graded critical, major, minor or info, with critical ones routed to a deviation, and the rules include absence (missing field, step, signature, or laboratory result). A review records positively whether issues were found, and the audit-trail reviewer is refused if they took part in the activities under review. QA approves the disposition first; QP certification is a separate, later signature. For Saudi manufacturers, see the notes on Corpex-EBR for SFDA-regulated sites.

None of that makes RBE defensible on its own. The validation evidence, the limits and the SOPs remain yours. The software only refuses to let you take the shortcut without them.

Frequently Asked Questions

Does EU GMP allow review by exception of batch records?

Not by name. Chapter 4 carries a note after paragraph 4.20 that where a validated process is continuously monitored and controlled, automatically generated reports may be limited to compliance summaries and exception or out-of-specification data reports. That note, together with Annex 15 on ongoing process verification and ICH Q10 on process performance monitoring, is the basis. It is conditional on the process being validated and continuously monitored.

Can a site use review by exception for products supplied to the US?

21 CFR 211.192 requires the quality control unit to review and approve all production and control records before release, and it contains no exception wording. A site can defend an exception-based procedure as the way that review is performed, but only with evidence that the system has verified the data not read line by line. Many sites keep full review as the default for US supply and permit RBE per product.

Does the QP still certify a batch released by exception?

Yes. Under Annex 16, batches are released only after certification by a QP, who is responsible for confirming that records are complete and that processes remain in the validated state. Review by exception changes how the records are reviewed before that point, not who certifies the batch.

Is the 2025 EU draft of Chapter 4 in force?

No. The revised Chapter 4, revised Annex 11 and new Annex 22 were drafts put to public consultation from 7 July to 7 October 2025. As of 11 November 2026 the European Commission still lists the 2011 versions of Chapter 4 and Annex 11 as current. Do not cite the drafts as requirements.

Sources

1. EU GMP Volume 4, Chapter 4: Documentation (January 2011), paragraph 4.20 and the note that follows it
2. EU GMP Annex 15: Qualification and Validation, section 5 (5.29, 5.31, 5.32)
3. EU GMP Annex 11: Computerised Systems, section 9
4. EU GMP Annex 16: Certification by a Qualified Person and Batch Release, 1.7.11, 1.7.12, 1.7.16 and 4.1
5. ICH Q10: Pharmaceutical Quality System, section 3.2.1
6. 21 CFR 211.188, Batch production and control records
7. 21 CFR 211.192, Production record review
8. European Commission: stakeholders consultation on Chapter 4, Annex 11 and Annex 22 (draft texts; consultation closed 7 October 2025)
9. European Commission: EudraLex Volume 4 index, checked 11 November 2026